On the Dual Effect of Bankruptcy

Transcription

1 On the Dual Effect of Bankruptcy Daiki Asanuma Abstract This paper examines whether the survival of low-productivity firms in Japan has prevented economic recovery since the bursting of the financial bubble in the late 1980s. The existence of so-called zombie firms is one reason for the stagnation of the Japanese economy, because they prevent more productive companies from gaining market share and thus reduce productivity gains for the overall economy. However, if the bankruptcy of one firm affects others in its network, this argument does not hold because networked firms can become embroiled in a bankruptcy chain. This paper assesses the validity of zombie theory using computer simulations within a network economy setting. It finds that governmental policies to save bankruptcy candidates improve macroeconomic performance in a network economy. In other words, governmental intervention can be effective in this kind of economy by preventing the propagation of a bankruptcy chain that may embroil high-performing firms. Keywords: Dual effect of bankruptcy, Network on firms, Agent based model JEL codes: E27, E69, H3 1. Introduction Since the bursting of the economic bubble in 1989, the Japanese economy has faced prolonged recession because of low productivity growth (Hayashi and Prescott, 2002). Recent studies such as Ahearne and Shinada (2005), Hoshi (2006), and Caballero et al. (2008) have all insisted that low productivity in Japan is the result of the existence of unprofitable and less productive firms, which have been termed zombie firms. 1 This paper examines whether such low-productivity firms should be encouraged to leave the market in order to improve overall macroeconomic performance. One line of reasoning suggests that firms that have high debt and low productivity should leave the market because they distort the normal functioning of market competition. If Japanese banks rationally decided to no longer lend to zombie firms, such companies would go bankrupt and, as a result, Japanese macroeconomic performance would improve. According to this argument, which concurs with neoclassical economic growth theories, the main engine for economic growth is Post-Doctoral Fellow, Graduate School of Economics and Management, Tohoku University Address: 27-1, Kawauchi, Aoba-ku, Sendai, Japan [email protected] 1 See Caballero et al. (2008).

2 total factor productivity. If a population desires high macroeconomic performance, it is necessary to increase productivity. However, the process of creative destruction, namely where new high-productivity market entrants actively replace old unproductive firms, does not necessarily work because of the existence of zombie firms 2. [Z]ombie firms prevent more productive companies from gaining market share, strangling a potentially important source of productivity gains for the overall economy (Ahearne and Shinada, 2005, p. 364). Because, by definition, the productivity of zombie firms is low, governments or banks must leave them to die. In other words, doing nothing is the best way to improve macroeconomic performance. This argument is based on the assumption that a firm s operations are independent of those of other firms. However, the standard method used to assess this situation, namely the representative agent model, cannot accurately model the interaction between heterogeneous agents. In particular, it overlooks the chain effect of bankruptcy (bankruptcy chain hereafter) through which the overall macroeconomic system can be affected. It is clear that agents in a macroeconomic system are connected with each other through various networks. In other words, they are dependent on each other. Indeed, when a firm goes bankrupt in the real world, the effect of this event may be propagated throughout its networks in the form of a bankruptcy chain, which is frequently seen during periods of economic depression. Therefore, it is difficult to distinguish the cause of bankruptcy because a firm can be involved in the bankruptcy of other firms in its network. This perspective contrasts with zombie theory, suggesting that the relationship between firm productivity and macroeconomic performance is complicated and thus that zombie theory does not always hold. If there does exist such a dual effect of bankruptcy, policy implications might also differ. In other words, certain firms should be given a lifeline (e.g., through the injection of external funds) even if they are considered to be zombies, because in a network economy firms that have relatively high productivity can nonetheless become embroiled in a bankruptcy chain. In Japan, government expenditure tends to be reduced from the perspective of the unprecedented high amount of government deficit. However, the reasons for this reduction have yet to be examined thoroughly. The present paper bridges this gap in the literature by using a computer simulation to analyze whether zombie theory holds in a network economy setting. We use the agent-based model proposed by Delli Gatti et al. (2005) in which there are a large number of heterogeneous interacting agents. The interaction between these heterogeneous agents generates emergence, which occurs when wholes (e.g., the economy) produce outcomes that differ categorically from those that the parts (e.g., human agents) can produce individually (Harper and Lewis, 2012, p. 329). 2 The most influential economist on the creative destruction is, of course, Schumpeter. He originally advanced the argument that recessions promote a more efficient allocation of resources by driving out bad investments and freeing up resources for more productive uses (Barlevy 2002, p.65).

3 Further, we consider a network to consist of financial contracts between each firm and a bank. In this sense, the explicit interconnection between firms cannot be detected because we do not introduce direct connections among them. Each firm, however, is connected implicitly through the variation in the conditions of financial contracts (i.e., interest rates charged by the bank). When a firm goes bankrupt, non-performing loans have to be wiped out by the same amount as the bank's net worth, which worsens the financial condition of the bank. Therefore, the bank raises interest rates to incumbent firms, which can be a trigger of the bankruptcy chain. In this model, therefore, the operations of all firms are dependent on those of others. The remainder of this paper is organized as follows. Section 2 describes the agent-based model. Because there are a large number of firms in this model, a simulation is carried out in Section 3. We shortly discuss limitations of this analysis in section 4. Section 5 concludes. 2. Model In this section, we build an agent-based model that models the offering of financial aid to low-productivity firms following Delli Gatti et al. (2005). 2.1 Firm Sector In the overall macroeconomic system, there are firms indexed by. The total number of firms is assumed to be fixed. Each firm has a simple production function,, where,, and represent output, productivity, and capital stock at time, respectively. We assume that firm s productivity is the summation of its basic level of productivity, denoted by, and a stochastic element. That is,, (1) where follows a uniform distribution that supports and. Therefore, the expectation value is time-invariant. On its balance sheet, firm possesses capital stock, liability stock, and net worth. The balance sheet condition requires that the following equation must hold for all :. (2) We ignore the demand side of the economy. This means that firm s sales are and it incurs costs of, where and represent the charged interest rate and rate of return, respectively. Assuming that the rate of return on net worth is equal to the charged interest rate and that costs other than interest repayments are proportional to interest repayments, profit can be

4 written as, where. (3) The present amount of net worth is composed of previous net worth and present profit. Therefore, net worth varies following the rule From equation (4), the level of productivity shock derived from (4) at which firm s net worth is negative can be (5) If, this firm s net worth is negative. In the case that there is no financial aid for firms that have negative net worth, such firms go bankrupt. The probability of becoming a bankruptcy candidate is determined by two financial variables: the interest rate charged to this firm and its own net worth. The higher the interest rate is and the smaller this firm s net worth is, the higher this probability becomes. Further, firms that have a high basic level of productivity have greater tolerance to productivity shocks. In other words, the lower the basic level of productivity is, the higher the probability of bankruptcy is. Following Delli Gatti et al. (2005), each firm maximizes its expected profit with the probability of becoming a bankruptcy candidate. Firm s objective function is thus. (6) By differentiating equation (6) with respect to, we can derive desired capital stock, that is,. (7) Each firm s investment is determined by the difference between desired capital stock and previous capital stock, namely. This investment is financed by the profit earned in previous periods and bank credit newly borrowed in the present period. Therefore,, (8) where denotes the credit demand to finance the necessary investment. Finally, we derive the credit demand function by rearranging equation (8) with the balance sheet condition and substituting equation (7) as follows:

5 . (9) 2.2 Government We assume that the government plays a role in helping bankruptcy candidates (i.e., zombies). It thus adheres to the following process: 1) At the end of each period t, the government gathers information on bankruptcy candidates. 2) Following particular policies (see Section 3), it chooses the candidates to be helped. 3) The government then allocates resources to each of them. The amount of resources is set as the same amount as the firm s net worth held just before it became a bankruptcy candidate. 4) The government raises these resources as a lump sum tax from aggregate production to save bankruptcy candidates. 2.3 Banking Sector The model of banking behavior also rests on Delli Gatti et al. s (2005) formalization, as no original view exists in the banking sector 3. We assume that the role of banks is only to create explicit and implicit financial connections through which external technological shocks can be propagated. At each time t, the bank s balance sheet thus becomes (10) where represent the stock of lending supply, deposits, and the bank s net worth, respectively. The bank faces a quantity constraint in terms of the total amount of credit supply, which is proportional to its net worth at period t with a constant :. (11) From the balance sheet constraint (10) and credit supply rule (11), we determine deposits is simply determined as the difference between credit supply and net worth. The bank thus supplies credit to each operating firm following this simple rule:, (12) where. Equation (12) states that the bank rations its total amount of credit depending on relative firm size, which is represented by the ratio of a firm s capital stock to total capital stock. From the credit demand function (9) and credit supply rule (12), the interest rate charged to firm is determined as follows: 3 Asanuma (forthcoming) which analyzes the relationship between bank lending attitudes and financial fragility provides a model based on the same framework with a view of post-keynesian banking sector.

6 . (13) The bank s profit is thus, (14) where and represent the average interest rate in period and a parameter that indicates the monopolistic power of the bank, respectively. In each period, firms may have negative net worth (i.e., be bankruptcy candidates). If they are not helped, they go bankrupt. The total amount of negative net worth then becomes the stock of non-performing loans, denotes a set of bankrupt firms. The bank wipes out these non-performing loans by subtracting them from its own net worth. As a result, the law of motion of a bank s net worth is (16) Based on this model setting, it is possible to replicate the process of bankruptcy chain. The mechanism is as follows; Because bankruptcy (increase of the stock of non-performing loans) always erodes the bank s financial condition through equation (16), this erosion results in the decrease of the capacity of credit supply as shown in equations (11) and (12). In this situation, incumbent firms must have harder financial contracts as higher interest rates and smaller amount of credits. Furthermore, the high interest repayment costs affect the firms profits negatively as shown in equation (3) and it increases the bankrupt probabilities of these firms. That is, bankruptcy in a particular period raises the bankrupt probability of the following periods. This is the mechanism of bankrupt chain in this kind of model. 3. Simulation For the computer simulation, the following parameters are used: number of firms, simulation span, maximum value of productivity, minimum value of productivity, bankruptcy cost coefficient total cost coefficient initial capital stock 1, initial net worth, initial borrowing, monopolistic power parameter, quantity constraint on lending capacity, initial lending, initial net worth of the bank, initial deposits, and the stochastic technical shock is subtracted from uniform distribution with the support [ 0.035, 0.035]. Further, we replicate the process of creative destruction in a straightforward way. The number of firms is fixed over the simulation period, which means that the number of exits (i.e., bankrupt firms) is equal to that of new entrants. Moreover, every firm has its own basic level of productivity. Therefore, if firm that has a basic level of productivity goes bankrupt, a

7 new firm enters that has the same number and basic level of productivity. However, these two basic levels of productivity are not the same number. We assume that the basic level of productivity for the new entrant is always 2% higher than that of the bankrupt firm. In other words, we set 4. As mentioned above, firms that have negative net worth are categorized as bankruptcy candidates. The following four policies are implemented to save such firms: 1) POLICY 1: Bankruptcy candidates that are in the top % of productivity are saved 2) POLICY 2: Bankruptcy candidates that are in the bottom % of productivity are saved 3) POLICY 3: No bankruptcy candidates are saved 4) POLICY 4: All bankruptcy candidates are saved In this simulation, is set as a parameter with a value of 3. Fig. 1 GDP dynamics 4 We totally ignore a job matching process pointed out by Mortensen and Pissarides (1994) and Barlevy (2002). Therefore, the new entrants are always operative.

8 Fig. 2 Bankruptcy rates Fig. 3 Non-performing loan rates

9 Fig. 4 Average productivities Fig. 1 shows the logarithmic GDP dynamics that result from this simulation and contains four lines: POLICY 1 corresponds to the blue line, POLICY 2 to the red line, POLICY 3 to the green line, and POLICY 4 to the brown line. Although all lines oscillate throughout the simulation period, the performance of POLICY 3 seems to be remarkably inferior to that of the other three policies. In POLICY 3, all bankruptcy candidates go bankrupt, which erodes the bank s financial condition by increasing non-performing loans. This increase results in a rise in the interest rates charged to operating firms in the next period (from equations (11) and (12)), which can trigger other bankruptcies because of the rise in the bankruptcy probability shown in equation (5). In brief, the economic system is unstable under POLICY 3. Figs. 2 and 3 support this view. Fig. 2 plots the bankruptcy rate, namely the ratio of bankrupt firms to incumbents. In the second half of the simulation period, that rate tends to be higher under POLICY 3 than it is under the other policies. Further, as shown in Fig. 3, the non-performing loan rate, namely the ratio of non-performing loans to total lending, is highest under POLICY 3 in almost all periods. Next, we move to the time evolution of average productivity (Fig. 4). Because we assume that the basic level of productivity of a new entrant is 2% higher than that of a bankrupt firm with the same number, the larger the number of bankruptcies is, the higher average productivity becomes. POLICY 3 realizes the highest average productivity in line with orthodox economics theory. We thus confirm that if the government does not intervene in the bankruptcy process, bankruptcy candidates smoothly exit the market and new entrants that have higher productivities continue to operate. As a result, average productivity in the macroeconomy increases compared with if the government were to intervene to save bankruptcy candidates.

10 However, as explained earlier, this result does not suggest that POLICY 3 realizes the highest macroeconomic performance because of the bankruptcy chain that could continue for several periods. Thus, we find a dual effect of bankruptcy. The first effect is the process of creative destruction, which is positive. However, because bankruptcy can be a trigger of other bankruptcies, a networked firm may become bankrupt simply because other firms have done so, which is negative. In summary, the presented results suggest that the relationship between average productivity and aggregate performance is less direct than orthodox economics implies. This novel contribution of the present paper thus points out the factors that orthodox economics ignores. Here, we can find the meaning of fiscal expenditures. From Fig. 1, POLICY 2 improves performance compared with the other policies. Because low-productivity firms face a high bankruptcy probability in our model, the government helps them, thereby allowing the creative destruction process to work and preventing a bankruptcy chain. We thus present a theoretical basis for these kinds of policies in contrast to the often proposed emotional argument. 4. Limitations In this section, we refer to the limitations of this analysis. First, the most point is to provide the empirical supports on parameters. As simulating the above model, we set some parameter values following Delli Gatti et al (2005) and set the others discretionally. We expect that this task would be hard. However, it is necessary to improve the robustness of this paper. Furthermore, we ignore some important economic structures, for example, the interbank market, other types of financial contracts, direct relationship among firms, and so on. These things reinforce the network structure in the macroeconomy. Therefore, the scope of our future research is to expand the presented model to include some factors we ignored here and try the empirical support. 5. Conclusion This paper examined the validity of zombie theory. Proponents of this theory insist that the existence of low-productivity firms in the market hinders the creative destruction process that allows the normal entry and exit mechanism to make the overall economy productive. The solution to improving such an economy is simple: encourage all zombies to leave the market. However, this argument implicitly assumes that all bankruptcies occur independently and do not affect other firms operations. When all firms are connected by an economic network, however, this argument does not hold. Using a computer simulation, this study confirmed that governmental policies to save bankruptcy candidates improve macroeconomic performance in a network economy. In other words, governmental intervention can be effective in this kind of economy by preventing the propagation of a bankruptcy chain that may embroil high-performing firms.

11 References Ahearne A. G. and N. Shinada, Zombie Firms and Economic Stagnation in Japan, International Economics and Economic Policy, Vol. 2, No. 4, pp: , Asanuma, D. Lending Attitudes as a Financial Accelerator in a Credit Network Economy, Journal of Economic Interaction and Coordination, Forthcoming. Barlevy, G. The Sullying Effect of Recessions, Review of Economic Studies, Vol. 69, No. 1, pp: 65-96, Caballero, R., T. Hoshi, and A. K. Kashyap, Zombie Lending and Depressed Restructuring in Japan, American Economic Review, Vol. 98, No. 5, pp: , Delli Gatti, D., C. D. Guilmi, E. Gaffeo, G. Giulioni, M. Gallegati, A. Palestrini, A New Approach to Business Fluctuations: Heterogeneous Interacting Agents, Scaling Laws and Financial Fragility, Journal of Economic Behavior & Organization, Vol. 56, No. 4, pp: , 2005 Harper, D. A, and P. Lewis, New Perspectives on Emergence in Economics, Journal of Economic Behavior & Organization, Vol. 82, No. 2-3, pp: , Hayashi, F. and E. C. Prescott, The 1990s in Japan: A Lost Decade, Review of Economic Dynamics, Vol. 5, No. 1, pp: , 2002.Hoshi, T. Economics of the living dead, Japanese Economic Review, Vol. 57, No. 1, pp: 30-49, Mortensen, D. T. and C. A. Pissarides, Job Creation and Job Destruction in the Theory of Unemployment, Review of Economic Studies, Vol. 61, No. 3, pp: , 1994.